Individual differences in EEG correlates of recognition memory due to DAT polymorphisms

Abstract Introduction Although previous research suggests that genetic variation in dopaminergic genes may affect recognition memory, the role dopamine transporter expression may have on the behavioral and EEG correlates of recognition memory has not been well established. Objectives The study aims to reveal how individual differences in dopaminergic functioning due to genetic variations in the dopamine transporter gene influences behavioral and EEG correlates of recognition memory. Methods Fifty‐eight participants performed an item recognition task. Participants were asked to retrieve 200 previously presented words while brain activity was recorded with EEG. Regions of interest were established in scalp locations associated with recognition memory. Mean ERP amplitudes and event‐related spectral perturbations when correctly remembering old items (hits) and recognizing new items (correct rejections) were compared as a function of dopamine transporter group. Results Participants in the dopamine transporter group that codes for increased dopamine transporter expression (10/10 homozygotes) display slower reaction times compared to participants in the dopamine transporter group associated with the expression of fewer dopamine transporters (9R‐carriers). 10/10 homozygotes further displayed differences in ERP and oscillatory activity compared to 9R‐carriers. 10/10 homozygotes fail to display the left parietal old/new effect, an ERP signature of recognition memory associated with the amount of information retrieved. 10/10 homozygotes also displayed greater decreases of alpha and beta oscillatory activity during item memory retrieval compared to 9R‐carriers. Conclusion Compared to 9R‐carriers, 10/10 homozygotes display slower hit and correct rejection reaction times, an absence of the left parietal old/new effect, and greater decreases in alpha and beta oscillatory activity during recognition memory. These results suggest that dopamine transporter polymorphisms influence recognition memory.


| INTRODUCTION
Recognition memory refers to an individual's ability to correctly identify previously encountered stimuli and is influenced by genetic variation in dopaminergic genes (Jocham et al., 2009;Li et al., 2013;Papassotiropoulos & Quervain, 2011;Papenberg et al., 2014;Schott et al., 2006;Takahashi et al., 2007). Specifically, altered dopamine transporter expression resulting from the dopamine transporter gene (DAT) affects behavioral and neuroimaging correlates of long-term memory processes (Li et al., 2013;Schott et al., 2006). The differential dopaminergic neurotransmission that results from the varied expression of DAT may alter the course of recognition memory retrieval processes, resulting in differences between individuals' ability in identifying previously encountered stimuli. However, it is currently unclear whether DAT genetic variation affects an individuals' recognition memory through processes associated with the retrieval of information itself or through cognitive control processes that serve to monitor and evaluate retrieved information. Therefore, this study uses electroencephalography (EEG) in combination with genetic data collection to show how dopaminergic transporter polymorphisms may alter the processes underlying memory retrieval during recognition memory.
Differences in recognition memory, ERP old/new effects, and oscillatory activity associated with recognition memory were examined between participants homozygous for the 10-repeat (10R) VNTR of the dopamine transporter gene and participants possessing a copy of the 9 (9R) repeat VNTR during an item memory task. Previous research suggests that decreased synaptic dopamine clearance is beneficial for memory (Li et al., 2013;Schott et al., 2006). Therefore, we hypothesize that participants homozygous for the 10R-allele, which results in increased DAT expression (Fuke et al., 2001) and increased dopaminergic clearance (Heinz et al., 2000) will display impaired item memory performance, alongside diminished ERP and oscillatory correlates of memory compared to participants that possess a 9R-allele.

| Participants
Seventy-six right handed participants from the University of Colorado Boulder community volunteered to participate in this study. All participants gave informed consent in accordance with the Institutional Review Board of the University of Colorado Boulder. Sixteen participants were removed from the study for various reasons. Four participants failed to complete the entirety of the study, and three were removed for technical reasons. Nine participants were removed due to excessive noise in the EEG recordings, including excessive blinking (n = 3), the required use of excessive channel interpolation (n = 2), the lack of 20 good hit and correct rejection epochs for comparison postartifact detection (n = 3), and the lack of adequate behavioral performance (n = 1). The removal of these participants resulted in a total of 60 participants aged 18-29 (mean ± standard deviation, 20.7 ± 2.59 years old; 27 females, 33 males) for analysis. DAT groups were split according to whether the variable nucleotide tandem repeat (VNTR) sequence that influences DAT expression repeated 9 or 10 times (Fuke et al., 2001). Of the 60 participants participating in the study, two participants (one male, one female) possessed a DAT genotype that failed to fit in either the established 9R-carrier or 10/10 homozygous group and were not included for DAT group analysis. The 31 participants that were heterozygous or homozygous for the 9Rallele (i.e., 9/9 or 9/10) were placed in one group (14 female, 17 male), whereas 27 participants (12 female, 15 male) homozygous for the 10R VNTR were placed in the other.

| Stimuli
Eight hundred and fifteen adjectives were used as stimuli. The Kucera and Francis (1967) word norms were used for the selection of adjectives in the study. The words were presented to the participants in white uppercase letters in the center of the screen on a 26 in LCD computer screen with a black background at a visual angle of 2.3° ( Figure 1). The average written frequency (kfreq) of all the adjectives used in the study was 34.86 and the average number of letters per word was 6.93. The average kfreq across the counterbalanced lists ranged from 34.19 to 35.93 and the average number of letters across counterbalanced lists ranged from 6.87 to 7.00 and the kfreq and number of letters did not differ between lists.

| Task
Participants performed an item memory task during one study session, and a separate, source memory task was performed during a separate study session on a different day. The source memory data will be presented elsewhere. For the item memory task, participants were presented a list of words and asked to encode them during the study phase. In order to familiarize participants with the task, participants first underwent a short practice block before being asked to encode words in the study block. During this practice block, participants were given instructions and studied 10 words in order to familiarize them with the task. Upon completion of the practice block, the study block began. The study block consisted of 204 words, with two words at the beginning and two words at the end of the study block acting as primacy and recency buffers. During the study block, participants were instructed to associate half of the words with the mental image of a place and the other half were asked to make a pleasantness rating (Davachi, Mitchell, & Wagner, 2003;Kahn, Davachi, & Wagner, 2004). A place or pleasantness cue was presented for 500 ms prior to adjective presentation, which lasted for 500 ms. A fixation cross was presented for 4,000 ms after adjective presentation to allow participants to perform the encoding task. Upon completion of the encoding period, a question mark popped up on the screen for 700 ms, a period in which participants were instructed to rate the degree to which they successfully encoded the adjective (Figure 1). Participants rated their performance by pressing one of four buttons: (1) unsuccessful; (2) partially successful; (3) successful with effort; (4) successful with ease.
Following the study block, item memory retrieval was tested while participants underwent EEG recording. Participants were fitted with a 128 channel Hydrocel Geodesic Sensor Net connected to an AC-coupled high input impedance amplifier (200 MΩ, Net Amps TM, Electrical Geodesics Inc., Eugene, OR). Amplified analog voltages (0.1-100 Hz bandpass) were digitized at 250 Hz. Individual sensors were adjusted until impedances were less than 50 kΩ. Participants were given a 15-word practice test block prior to beginning the retrieval task. Approximately 30 min passed between the conclusion of the encoding phase and the beginning of the retrieval phase of the study. Participants viewed 480 words during the item retrieval test: 200 previously studied words, 200 new words, and 80 words serving as buffers. The adjectives were presented in blocks of 24, with two words at the beginning and end of each block serving as primacy and recency buffers. Twenty test blocks were used to test item memory retrieval. For each presented adjective, there was an initial variable fixation period of 50-150 ms, followed by the test word for 750 ms and an additional fixation period of 1,750 ms. Participants were permitted to respond upon word presentation. To respond, participants used the index fingers of both hands and pressed one key for an old (previously studied word) and another key for a new word. Following their response, participants used the index and middle finger of one hand and the index finger of their other hand to provide information regarding the degree of confidence of their answer. One key was pressed for "surely," one key was pressed for "likely," and another key was pressed F I G U R E 1 Behavioral paradigm used during recognition memory task. During the encoding phase, participants were given a place or pleasantness cue for 500 ms indicating the task to use during encoding. Following this cue, an adjective was presented for 500 ms. Participants were given 4,000 ms to perform the encoding task and then were asked to rate how successfully they were performing the task. The bottom panel represents the retrieval phase where EEG recordings took place. A variable duration fixation cue was presented for 50-150 ms followed by an adjective for 750 ms and a fixation cross for 1,750 ms. Participants could respond at any time after presentation of the adjective with one of two choices, "new" or "old"

Item Retrieval
Response Types "new" "old" for "maybe." EEG data, accuracy data, and reaction time (RT) data were collected as participants completed the task.

| ERP preprocessing
For ERP preprocessing, EEGLAB (Delorme & Makeig, 2004; RRID: SCR_007292) and ERPLAB (Lopez-Calderon & Luck, 2014; RRID: SCR_009574) were used. Before data preprocessing, channels with excessive noise were identified via visual inspection and interpolated using spherical spline interpolation. Two participants that required the interpolation of more than five channels (4%) were not included in the final data analysis. Data processing included filtering the data from 0.1 to 40 Hz, rereferencing to the average signal, separating the data into epochs, and artifact rejection. The data were epoched into periods 800 ms prestimulus presentation to 1,500 ms poststimulus presentation (−800 to 1,500 ms). Epochs were sorted into bins according to their response type (hits and correct rejections). Correctly remember- ing an item as one previously encountered constituted a hit, whereas correctly indicating that a word had never been seen before constituted a correct rejection (CR). Artifact rejection was accomplished with an automated moving window search procedure where changes of 100 μV were marked for rejection in 50 ms bins of 100 ms length.
A threshold of 20 clean, artifact free epochs for each type of response

| Spectral analysis preprocessing
Spectral analyses were run to examine oscillatory power during hits and correct rejections. For the spectral analyses, datasets for item memory were repreprocessed in EEGLAB. Repreprocessing was done due to the differences in standard preprocessing steps for ERP and oscillatory analyses, particularly the need to use the cleanest data possible for oscillatory analysis. Preprocessing included filtering the data from 1 to 100 Hz, rereferencing to the average signal, and artifact rejection. Data were epoched into the same −800 to 1,500 ms epochs as for the ERP analysis and sorted into hits and correct re-

| Genotyping
Genomic DNA was isolated from saliva samples collected using a commercial product (Oragene ™ , DNAgenotek, Ottawa, ON, Canada).
A common genetic variant of the DAT gene (SLC6A3) is a 40-bp variable number tandem repeat (VNTR) sequence that repeats 9 or 10 times (Vandenbergh et al., 1992), with individuals possessing a copy of the 9-repeat VNTR (9-carriers) displaying decreased DAT expression (Fuke et al., 2001;Heinz et al., 2000) and increased synaptic dopamine (Heinz et al., 2000) compared to 10/10 homozygotes. This 40 bp DAT1 VNTR (rs28363170) was genotyped as described in Haberstick et al. (2014). During genotyping, roughly one-third of the samples (18 random, six for one or more genotype assignments) were regenotyped (a new PCR and fragment analysis) resulting in two previously failed samples to be assigned genotypes. All other samples were consistent between runs. DAT groups were split according to whether the variable nucleotide tandem repeat (VNTR) sequence that influences DAT expression repeated 9 or 10 times (Fuke et al., 2001). Participants that were heterozygous or homozygous for the 9-repeat version of the allele (i.e., 9/9 or 9/10) were placed in one group, whereas participants homozygous for the 10-repeat VNTR were placed in the other. The DAT genotype frequencies were distributed according to the Hardy-Weinberg Equilibrium (9.1% 9/9, 42.1% 9/10, 48.8% 10/10). EEGLAB's permutation-based statistics function was utilized and set to 1,000 permutations, and the p-value for statistical significance was set to p = .05 using an FDR correction for multiple comparisons.

| Parietal old/new effect during item memory
The 2 (hits vs. correct rejections) × 2 (DAT group) repeated measures ANOVA conducted for the left posterior superior ROI at 500-800 ms poststimulus presentation revealed a trend toward a significant main effect of condition (F 1,56 = 3.71, p = .06, partial η 2 = 0.06), and a significant interaction between condition and DAT group (F 1,56 = 4.86, p = .03, partial η 2 = 0.08  Figure 4b). Independent sample t tests indicated that there was no difference in CR mean amplitude between 10/10 homozygotes and 9R-carriers (t 56 = 0.65, p = .52, Cohen's d = 0.17, Figure 5). A trend toward a significant difference was observed when mean amplitude during hits was compared between DAT groups (t 56 = 1.89, p = .07, Cohen's d = 0.49, Figure 5) suggesting that 10/10 homozygous participants do not show the left parietal old/new effect due to decreased mean ERP amplitudes during hit trials compared to participants carrying a 9R-allele.

| Late posterior negativity during item memory
The late posterior negativity during the item memory task was ex-

| Effects of DAT polymorphism on oscillatory power during item memory
Postcomponent clustering, three distinct component clusters located in the midparietal region (40 participants, 80 independent components), midfrontal region (40 participants, 90 independent components), and left parietal region (34 participants, 57 independent components) displayed significant differences in oscillatory activity as a function of the DAT gene. The midparietal component cluster ( Figure 6) showed a significant effect of DAT group on oscillatory power for both hits and correct rejections. Significant differences in hit oscillatory power were observed between 10/10 homozygotes and 9-carriers in a frequency range from theta to early beta (5-18 Hz). Differences in theta band activity (4-8 Hz) were observed occurring from 740 to 1,108 ms poststimulus, whereas differences in alpha (8-12 Hz) and early beta band (13-18 Hz) activity were observed from roughly 740-1,218 ms postcue presentation. Analyses of correct rejection oscillatory activity revealed similar results, with significant differences in correct rejection oscillatory power observed between 10/10 homozygotes and 9-carriers in a frequency range from theta to beta (6.5-26 Hz).
Differences in theta band activity (6.5-8 Hz) were observed starting 845-1,218 ms postcue presentation, whereas differences in alpha in theta to beta (5-24 Hz) oscillatory power. A brief period of significantly different theta activity (5-8 Hz) was observed occurring from 741 to 967 ms. Differences in alpha and beta oscillatory activity (8-24 Hz) between 10/10 homozygotes and 9-carriers during item memory hits were observed to last for a longer period, with alpha band activity (8-12 Hz) significantly different from a period F I G U R E 5 Bar graph illustrating ERP amplitude differences in LPS 500-800 ms post-stimulus presentation during item memory. The standard error of the means are designated with error bars. Average ERP amplitudes for 10/10 homozygotes (10/10 hits, blue) and 9-carriers (9-carrier hits, purple) during item memory hits. Correct rejections are represented in orange for 10/10 homozygotes and in green for 9-carriers. The ERP amplitude between 10/10 homozygote and 9-carrier hits suggests a trend toward a significant difference, with 10/10 homozygotes displaying decreased ERP hit amplitude

| DAT polymorphism and behavioral correlates of item memory retrieval
Our study results show significant differences in response times as a function of DAT polymorphism, with participants homozygous for the 10R VNTR displaying significantly slower reaction times for both hits and correct rejections compared to participants possessing a 9R allele. Increasing dopamine levels during item memory results in faster response times for both hits and correct rejections (Apitz & Bunzeck, 2013;Bunzeck et al., 2009;Eckart & Bunzeck, 2013), demonstrating a link between dopamine and recognition response times. Dopaminergic neurotransmission is partially regulated by the dopamine transporter, which serves to retrieve synaptic dopamine and return it to presynaptic neurons, terminating dopaminergic signaling. Participants homozygous for the 10R VNTR have increased DAT expression (Fuke et al., 2001), increased dopaminergic reuptake, and decreased synaptic dopamine (Heinz et al., 2000). With studies showing that altering dopamine levels changes recognition response time (Apitz & Bunzeck, 2013;Bunzeck et al., 2009;Eckart & Bunzeck, 2013), the decreased synaptic dopamine hypothesized to be occurring in 10/10 homozygotes may explain the increased hit and correct rejection reaction times displayed by 10/10 homozygous participants performing our item memory task.

| DAT polymorphism and ERP correlates of recognition memory
Dopamine may contribute to recognition memory by modulating the level of information retrieved for postretrieval processing, with the dopamine transporter polymorphism playing a significant role in this modulation. ERP studies of recognition memory have identified several distinct neural correlates known as old/new effects, which are more positive ERP deflections for hits compared to correct rejections (Curran, 2000;Curran et al., 2006;Donaldson & Rugg, 1998;Friedman & Johnson, 2000;Rugg & Curran, 2007;Rugg et al., 1998;Vilberg & Rugg, 2007;Wilding & Rugg, 1996;Woodruff, Hayama, & Rugg, 2006). The left parietal old/new effect is associated with the amount of information retrieved (Vilberg & Rugg, 2007;Vilberg et al., 2006;Wilding, 2000). Specifically, Wilding (2000) show that the left parietal old/new effect is larger when retrieving more contextual details associated with an item. Additionally, Vilberg et al. (2006) found that the magnitude of the parietal old/new effect is larger when participants fully recollect available visual information compared to partial recollection using a Remember/Know task. Our results show that participants carrying a 9R-allele display a robust left parietal/old new effect, along with greater mean hit amplitude than their 10/10 counterparts ( Figure 4). Thus, 10/10 homozygotes may have access to less information during the item memory task resulting in slowed performance. Combined with the finding that 10/10 homozygous participants have increased dopamine transporter expression (Fuke et al., 2001), which is associated with increased synaptic dopamine clearance (Heinz et al., 2000), our results suggest dopamine function may relate to controlling the amount of information available during recognition memory.
Though the 10/10 homozygous participants do not show a left parietal old/new effect and are slowed during task performance, they still are accurate at identifying items as old or new. Accuracy in the recognition memory task may be related to other ERP components associated with memory. Our study results show that the item memory task elicited no significant differences in the early frontal or late frontal old/new effects across participants, regardless of DAT polymorphic group. Due to the presence of a main effect of condition during the late posterior negativity, subsequent analyses were further conducted for this later ERP signature. Subsequent analyses on the late posterior negativity, a hypothesized ERP signature of evaluative cognitive control processes associated with retrieved contextual details (Johansson & Mecklinger, 2003), showed that the LPN was present in both DAT groups, suggesting that the LPN occurs during the item memory task regardless of DAT polymorphism. This pattern of ERP results may provide some rationale as to why accuracy is unaffected, whereas mean reaction times were affected by DAT polymorphism. Alongside absent early frontal and left parietal old/new effects, these results suggest the LPN component may be associated with item memory performance in 10/10 homozygotes.

| DAT polymorphism and oscillatory correlates of item memory retrieval
Analyses of oscillatory activity in midparietal, midfrontal, and left pa- Increased alpha and beta desynchronization in 10/10 homozygotes may allow for accurate recognition of new or old items despite the lack of a parietal old/new effect. Alterations in alpha and beta power have been linked to memory processes (Fell et al., 2008;Fellner et al., 2013;Hanslmayr et al., 2009Hanslmayr et al., , 2011Sederberg et al., 2007;Waldhauser et al., 2012;Weiss & Rappelsberger, 2000). Specifically, the desynchronization hypothesis postulates that decreases in alpha and beta power, resulting in desynchronization of neural ensembles, are related to memory retrieval (Düzel et al., 2003;Hanslmayr et al., 2012;Khader & Rösler, 2011;Spitzer, Hanslmayr, Opitz, Mecklinger, & Bäuml, 2008), with larger decreases in alpha and beta power associated with the retrieval of more information (Khader & Rösler, 2011).
Neurons that fire synchronously convey less information compared to neurons that fire asynchronously Schneidman et al., 2011). Therefore, alpha and beta desynchrony may allow for a small network of neurons to generate an infinite number of neural firing patterns allowing a vast amount of information to be sent from a local neural assembly . Participants homozygous for the 10-repeat allele display significantly decreased alpha/beta power during hits in midparietal, midfrontal, and left parietal component clusters (Figures 6 and 7), with these decreases lasting for a longer period of time compared to 9R-carriers. These decreases of alpha/beta power coincide with prior results showing decreases in alpha/beta power during memory retrieval (Düzel et al., 2003;Khader & Rösler, 2011;Spitzer et al., 2008 Sauseng et al., 2006), decreased theta oscillatory activity over posterior parietal brain regions has also been related to attention, with a recent study performed by Friese et al. (2016) showing increased theta desynchronization when participants were required to attend to stimuli. Our study reveals significant differences in theta power between 9-carriers and 10/10 homozygotes in a midparietal component cluster during a memory retrieval task, a result that suggests that DAT genetic polymorphisms affect attentional processes underlying successful memory retrieval. The presence of decreased theta power in the midparietal component cluster for 10/10 homozygotes for both hits and correct rejections suggests that 10/10 homozygotes may be utilizing increased top-down or bottom-up attentional processes to properly identify items as new or old.

| Limitations
Our current study describes how individual differences in recognition memory are affected by genetic variation in the dopamine transporter gene by analyzing the behavioral, ERP, and oscillatory correlates of item memory retrieval. As we were focused on memory retrieval, no EEG data during the encoding process were recorded. Therefore, we cannot rule out that differences during encoding between our DAT groups could explain the differential EEG and behavioral results we observed. Future studies examining encoding differences should explore this issue. Additionally, Chabris et al. (2012) conclude that studies attempting to establish relationships between SNPs (single nucleotide polymorphisms) and cognitive abilities may be underpowered, requiring large participant numbers. Our study utilizes a relatively small sample of 58 participants, and this small sample size may account for our inability to find significant differences in item memory accuracy between our 9R-carrier and 10/10 homozygote groups.
However, our study displays moderate to large effect sizes regard-

| CONCLUSION
Our study aims to further understand individual differences in recognition memory by describing the effect dopamine transporter genetic variation has on both behavioral and electrophysiological correlates of recognition memory. Our results show that dopamine transporter genetic variation affects mean ERP amplitudes over left parietal scalp locations, with 10/10 homozygotes, who show increased DAT expression (Fuke et al., 2001), showing no left parietal old/new effect alongside significantly increased reaction times for hits and correct associated with the amount of information retrieved (Vilberg & Rugg, 2007;Vilberg et al., 2006;Wilding, 2000) and decreases in alpha and beta power have been associated with the increased transmission of information Khader & Rösler, 2011).
Therefore, our study suggests that individuals who have increased dopamine transporter expression may rely on the increased transmis-behavioral analyses, processed the ERP and oscillatory data, and wrote the manuscript. Robert Ross aided Paolo Medrano with analyzing the behavioral and EEG data, along with editing the manuscript.